In our research we need to apply a sequence of acoustic stimuli (24bit/96KHz or 24bit/192KHz) in pseudo-random order. These stimuli can be since single clicks (5ms-long while noise) up to 10s-long natural sounds. It's pretty important a precise signalling (<1ms) of the beginning and end of each stimulus presentation by using a TTL pulse. The HifiBerry's RCA output will be plugged into a input speaker's driver, which has been designed by TDT.
Based my first reading, once the HifiBerry is properly configured, the Linux on Raspberry will see it as audio device, which is usually directly accessible by using some /dev (/dev/dsp for instance). Therefore, to do that, we're considering to write a piece of C-code for Raspberry Pi B+, plugged to HifiBerry. The code will store the stimuli in a set of buffers (RAM), each stimulus in a buffer, and then play them according the given pseudo-random order, with proper TTL signal to mark the beginning and end of each stimulus presentation.
Does this make sense ?